People have found a new job for AI coding agents: stripping the “smart” features back out of their smart TVs. According to TechCrunch AI, developers and hobbyist vibe coders are now sharing prompts and GitHub projects that use tools like Claude Code to remove sponsored ads, disable viewing trackers, replace bloated launchers and speed up sluggish home screens. Some of these posts have gone viral, and it shows something bigger than a weekend hack.
What’s actually happening
TechCrunch AI highlights two projects in particular:
- dumbtv: Its GitHub page says, “Your TV should do what you paid for: your apps, your HDMI inputs, and nothing else.” Developer Muzzammil Yasin Sabuwala built it with a mix of Deepseek V4.1 Flash, Claude Opus 5.5 and Claude Sonnet 5.5. It’s been tested on Fire TV Edition sets and on Android/Google TV models from Sony, TCL, Philips, Hisense, Xiaomi and Nvidia Shield. You use it by pointing your own AI agent at the repo.
- cleanyourtv: This one removes bloatware from Android TV and Google TV over Wi-Fi and swaps in a cleaner, Apple TV-style launcher. It doesn’t need root access, and you can still factory reset afterward. Creator Gathin T. says he posted his prompt after fixing his parents’ TV, and “a few thousand people did the same to their own TVs over the next two days.”
Some people don’t use a repo at all. Ryan Doyle sent Claude Code his TV’s product listing and asked, “Can you hack this?” Mihir Gupta reports that one afternoon got him “25 bloat apps disabled” and a 192 MB stock launcher replaced with an 83 MB custom one.
Why this matters now
People have hacked their gadgets for decades. What’s new is who can do it. Getting the same result used to mean knowing ADB commands, package names and which system apps you could safely kill. Now someone with no technical background can describe the outcome they want and have an agent walk them through it.
The target makes this more pointed. Modern smart TVs use automatic content recognition (ACR), which fingerprints whatever is on screen through screenshots or audio samples. That data then feeds ad targeting and gets sold to third parties. For many TV makers, these platform and ad businesses earn more than the hardware does. Cheap TVs are often subsidized by monetizing what you watch.
TechCrunch AI suggests that if these hacks spread to everyday consumers, they could disrupt the smart TV ad business. That’s a reasonable read, but it’s an early one. Right now this is a viral moment among tech enthusiasts, not a mass movement.
The risks are real
The article is clear about the downsides:
- Claude sometimes refused simple “hack my TV” prompts, so users needed more detailed instructions.
- AI agents can disable critical system packages by mistake.
- At least one user says they bricked an Android TV, and another got stuck partway through modifying a Fire TV.
All of these projects are still works in progress. If you’d rather not risk it, established launcher replacements like Projectivy Launcher, ATV Launcher and FLauncher handle a lot of the cleanup with less danger.
Where this goes in the next 1-3 years
I’d watch three things:
- Manufacturers lock down. Expect firmware updates that restrict ADB access, block alternative launchers or limit sideloading. Device makers have done this before every time owners got too much control, and AI just raises the stakes.
- AI labs get pulled into the fight. If agents become the main way people modify devices they own, expect pressure on model providers over what their agents will and won’t do. The refusals people are already hitting show where that line is still unsettled.
- “Dumb” becomes a product feature. Some brand will likely market an ad-free, no-ACR TV as a premium option, much as privacy became a selling point for phones.
The bigger pattern is that coding agents let ordinary owners renegotiate their deals with device makers. TVs are first because the ads are so annoying. Cars, appliances and routers could easily be next.
What to do with this
- Consumers: Turn off ACR in your TV’s privacy settings first. It’s free and safe. If you try an AI-driven cleanup, check that you can roll it back before you start.
- Businesses in ad-supported hardware: Plan for some users to opt out. A revenue model that depends on owners never touching their settings is now a weak one.
- Developers: Projects like these work best when they’re documented, reversible and tested on specific devices. The ones that last will put safety ahead of speed.
The full story, including the viral before-and-after demos, is available at TechCrunch AI.